Skip to content

adobe.md

Latest commit

 

History

History
129 lines (102 loc) · 4.61 KB

adobe.md

File metadata and controls

129 lines (102 loc) · 4.61 KB

Challenge 1a: PDF Processing Solution

Overview

This is a sample solution for Challenge 1a of the Adobe India Hackathon 2025. The challenge requires implementing a PDF processing solution that extracts structured data from PDF documents and outputs JSON files. The solution must be containerized using Docker and meet specific performance and resource constraints.

Official Challenge Guidelines

Submission Requirements

  • GitHub Project: Complete code repository with working solution
  • Dockerfile: Must be present in the root directory and functional
  • README.md: Documentation explaining the solution, models, and libraries used

Build Command

docker build --platform linux/amd64 -t <reponame.someidentifier> .

Run Command

docker run --rm -v $(pwd)/input:/app/input:ro -v $(pwd)/output/repoidentifier/:/app/output --network none <reponame.someidentifier>

Critical Constraints

  • Execution Time: ≤ 10 seconds for a 50-page PDF
  • Model Size: ≤ 200MB (if using ML models)
  • Network: No internet access allowed during runtime execution
  • Runtime: Must run on CPU (amd64) with 8 CPUs and 16 GB RAM
  • Architecture: Must work on AMD64, not ARM-specific

Key Requirements

  • Automatic Processing: Process all PDFs from /app/input directory
  • Output Format: Generate filename.json for each filename.pdf
  • Input Directory: Read-only access only
  • Open Source: All libraries, models, and tools must be open source
  • Cross-Platform: Test on both simple and complex PDFs

Sample Solution Structure

Challenge_1a/
├── sample_dataset/
│   ├── outputs/         # JSON files provided as outputs.
│   ├── pdfs/            # Input PDF files
│   └── schema/          # Output schema definition
│       └── output_schema.json
├── Dockerfile           # Docker container configuration
├── process_pdfs.py      # Sample processing script
└── README.md           # This file

Sample Implementation

Current Sample Solution

The provided process_pdfs.py is a basic sample that demonstrates:

  • PDF file scanning from input directory
  • Dummy JSON data generation
  • Output file creation in the specified format

Note: This is a placeholder implementation using dummy data. A real solution would need to:

  • Implement actual PDF text extraction
  • Parse document structure and hierarchy
  • Generate meaningful JSON output based on content analysis

Sample Processing Script (process_pdfs.py)

# Current sample implementation
def process_pdfs():
    input_dir = Path("/app/input")
    output_dir = Path("/app/output")
    
    # Process all PDF files
    for pdf_file in input_dir.glob("*.pdf"):
        # Generate structured JSON output
        # (Current implementation uses dummy data)
        output_file = output_dir / f"{pdf_file.stem}.json"
        # Save JSON output

Sample Docker Configuration

FROM --platform=linux/amd64 python:3.10
WORKDIR /app
COPY process_pdfs.py .
CMD ["python", "process_pdfs.py"]

Expected Output Format

Required JSON Structure

Each PDF should generate a corresponding JSON file that must conform to the schema defined in sample_dataset/schema/output_schema.json.

Implementation Guidelines

Performance Considerations

  • Memory Management: Efficient handling of large PDFs
  • Processing Speed: Optimize for sub-10-second execution
  • Resource Usage: Stay within 16GB RAM constraint
  • CPU Utilization: Efficient use of 8 CPU cores

Testing Strategy

  • Simple PDFs: Test with basic PDF documents
  • Complex PDFs: Test with multi-column layouts, images, tables
  • Large PDFs: Verify 50-page processing within time limit

Testing Your Solution

Local Testing

# Build the Docker image
docker build --platform linux/amd64 -t pdf-processor .

# Test with sample data
docker run --rm -v $(pwd)/sample_dataset/pdfs:/app/input:ro -v $(pwd)/sample_dataset/outputs:/app/output --network none pdf-processor

Validation Checklist

  • All PDFs in input directory are processed
  • JSON output files are generated for each PDF
  • Output format matches required structure
  • Output conforms to schema in sample_dataset/schema/output_schema.json
  • Processing completes within 10 seconds for 50-page PDFs
  • Solution works without internet access
  • Memory usage stays within 16GB limit
  • Compatible with AMD64 architecture

Important: This is a sample implementation. Participants should develop their own solutions that meet all the official challenge requirements and constraints.